The approach described in this paper: having an archeologist implement a virtual machine in an afternoon, and then having that virtual machine run a self-checking program that expanded to a full system, is exactly what contestants/archeologists in ICFP 2006 did.
I've been returning to the contest recently. There's a treasure trove of interesting stuff, including the ML compiler that targets the universal machine. It's amazing that such a small group of people created all that for a programming contest.
http://www.boundvariable.org
I've been returning to the contest recently. There's a treasure trove of interesting stuff, including the ML compiler that targets the universal machine. It's amazing that such a small group of people created all that for a programming contest.